We will determine whether kallikrein forms (prostate-specific antigen [PSA] and kallikrein 2 [hK2]) measured in men aged 44 - 50 predict subsequent metastases or death from prostate cancer. No previous study have evaluated whether a biomarker analyzed in a healthy volunteer population predicts metastases or cancer-specific death many years (typically 15-35 years) later. Low incidence of PSA testing in Sweden during enrollment (1974-1986) and long-term follow-up are prerequisites for our unique "natural experiment". We do not ask "can we detect prostate cancer?" but "can we predict prostate cancers with important impact on quality of life or reduced life expectancy?" We wish to predict risk of metastases or death from prostate cancer to risk-stratify screening and chemoprevention strategies, which must be made before these efforts start;i.e. in age 44- 50. We will first determine whether highly optimized measures of PSA and hK2 in blood collected at age =50 predict later development of metastases or death from prostate cancer;go on to determine whether measures in a 2nd sample (collected 6 years later) enhances these predictions. We will build statistical models to predict an individual man's probability of metastases or death from prostate cancer using single, or repeated measures at ages 44 - 55. If these models have sufficient accuracy in the R21 phase, we will go on in the R33 phase to determine their accuracy on an independent cohort of 40,000 men. Many common problems are avoided. These include over-diagnosis: we know that PSA detects prostate cancer at early, curable stages with high sensitivity (but poor specificity), however, many PSA-detected cancers would never cause morbidity or mortality to the patient before he dies of other causes. Studies are also affected by selection bias: men with family history or other risk factors are more likely to attend screening;and lead-time bias, where apparent survival advantage only relates to early time of diagnosis. Regarding the key issue of over-diagnosis, our study endpoint is death, or metastatic disease. There will be little if any selection bias due to low incidence of PSA screening in our study cohorts. Due to the retrospective nature of the study, the participants are never influenced by any PSA-data, and we will not be subject to lead- time bias as we will not compare survival between screened vs. unscreened cohorts. Using blood samples taken in Sweden in the mid-1970s - 1980s, we were able to show that a single blood test at age 44 - 50 can predict prostate cancer up to 25 years later. As many more men are diagnosed with prostate cancer than who die from the disease, the current proposal is to extend our work to determine if we can predict prostate cancer death. If so, a single blood test could be used to identify men at highest risk from prostate cancer and particular efforts could then be made for intensive screening and chemoprevention for these individuals.